Vehicle Detection Project¶
Cell 1: All the import statements needed for this project¶
In [1]:
import matplotlib.image as mpimg
import matplotlib.pyplot as plt
import numpy as np
import cv2
import glob
import time
from sklearn.svm import LinearSVC
from sklearn.preprocessing import StandardScaler
from skimage.feature import hog
from lesson_functions import *
# NOTE: the next import is only valid for scikit-learn version <= 0.17
# for scikit-learn >= 0.18 use:
from sklearn.model_selection import train_test_split
# from sklearn.cross_validation import train_test_split
import pickle
Cell 2: Util functions¶
In [2]:
def convert_color(img, conv='RGB2YCrCb'):
if conv == 'RGB2YCrCb':
return cv2.cvtColor(img, cv2.COLOR_RGB2YCrCb)
if conv == 'BGR2YCrCb':
return cv2.cvtColor(img, cv2.COLOR_BGR2YCrCb)
if conv == 'RGB2LUV':
return cv2.cvtColor(img, cv2.COLOR_RGB2LUV)
if conv == 'RGB2YUV':
return cv2.cvtColor(img, cv2.COLOR_RGB2YUV)
# Define a function to return HOG features and visualization
def get_hog_features(img, orient, pix_per_cell, cell_per_block,
vis=False, feature_vec=True):
# Call with two outputs if vis==True
if vis == True:
features, hog_image = hog(img, orientations=orient,
pixels_per_cell=(pix_per_cell, pix_per_cell),
block_norm= 'L2-Hys',
cells_per_block=(cell_per_block, cell_per_block),
transform_sqrt=True,
visualize=vis, feature_vector=feature_vec)
return features, hog_image
# Otherwise call with one output
else:
features = hog(img, orientations=orient,
pixels_per_cell=(pix_per_cell, pix_per_cell),
cells_per_block=(cell_per_block, cell_per_block),
block_norm= 'L2-Hys',
transform_sqrt=True,
visualize=vis, feature_vector=feature_vec)
return features
# Define a function to compute binned color features
def bin_spatial(img, size=(32, 32)):
# Use cv2.resize().ravel() to create the feature vector
features = cv2.resize(img, size).ravel()
# Return the feature vector
return features
# Define a function to compute color histogram features
# NEED TO CHANGE bins_range if reading .png files with mpimg!
def color_hist(img, nbins=32, bins_range=(0, 256)):
# Compute the histogram of the color channels separately
channel1_hist = np.histogram(img[:,:,0], bins=nbins, range=bins_range)
channel2_hist = np.histogram(img[:,:,1], bins=nbins, range=bins_range)
channel3_hist = np.histogram(img[:,:,2], bins=nbins, range=bins_range)
# Concatenate the histograms into a single feature vector
hist_features = np.concatenate((channel1_hist[0], channel2_hist[0], channel3_hist[0]))
# Return the individual histograms, bin_centers and feature vector
return hist_features
# Define a function to extract features from a list of images
# Have this function call bin_spatial() and color_hist()
def extract_features(imgs, color_space='RGB', spatial_size=(32, 32),
hist_bins=32, orient=9,
pix_per_cell=8, cell_per_block=2, hog_channel=0,
spatial_feat=True, hist_feat=True, hog_feat=True):
# Create a list to append feature vectors to
features = []
# Iterate through the list of images
for file in imgs:
file_features = []
# Read in each one by one
image = mpimg.imread(file)
# apply color conversion if other than 'RGB'
if color_space != 'RGB':
if color_space == 'HSV':
feature_image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV)
elif color_space == 'LUV':
feature_image = cv2.cvtColor(image, cv2.COLOR_RGB2LUV)
elif color_space == 'HLS':
feature_image = cv2.cvtColor(image, cv2.COLOR_RGB2HLS)
elif color_space == 'YUV':
feature_image = cv2.cvtColor(image, cv2.COLOR_RGB2YUV)
elif color_space == 'YCrCb':
feature_image = cv2.cvtColor(image, cv2.COLOR_RGB2YCrCb)
else:
feature_image = np.copy(image)
if spatial_feat == True:
spatial_features = bin_spatial(feature_image, size=spatial_size)
file_features.append(spatial_features)
if hist_feat == True:
# Apply color_hist()
hist_features = color_hist(feature_image, nbins=hist_bins)
file_features.append(hist_features)
if hog_feat == True:
# Call get_hog_features() with vis=False, feature_vec=True
if hog_channel == 'ALL':
hog_features = []
for channel in range(feature_image.shape[2]):
hog_features.append(get_hog_features(feature_image[:,:,channel],
orient, pix_per_cell, cell_per_block,
vis=False, feature_vec=True))
hog_features = np.ravel(hog_features)
else:
hog_features = get_hog_features(feature_image[:,:,hog_channel], orient,
pix_per_cell, cell_per_block, vis=False, feature_vec=True)
# Append the new feature vector to the features list
file_features.append(hog_features)
features.append(np.concatenate(file_features))
# Return list of feature vectors
return features
# Define a function that takes an image,
# start and stop positions in both x and y,
# window size (x and y dimensions),
# and overlap fraction (for both x and y)
def slide_window(img, x_start_stop=[None, None], y_start_stop=[None, None],
xy_window=(64, 64), xy_overlap=(0.5, 0.5)):
# If x and/or y start/stop positions not defined, set to image size
if x_start_stop[0] == None:
x_start_stop[0] = 0
if x_start_stop[1] == None:
x_start_stop[1] = img.shape[1]
if y_start_stop[0] == None:
y_start_stop[0] = 0
if y_start_stop[1] == None:
y_start_stop[1] = img.shape[0]
# Compute the span of the region to be searched
xspan = x_start_stop[1] - x_start_stop[0]
yspan = y_start_stop[1] - y_start_stop[0]
# Compute the number of pixels per step in x/y
nx_pix_per_step = np.int(xy_window[0]*(1 - xy_overlap[0]))
ny_pix_per_step = np.int(xy_window[1]*(1 - xy_overlap[1]))
# Compute the number of windows in x/y
nx_buffer = np.int(xy_window[0]*(xy_overlap[0]))
ny_buffer = np.int(xy_window[1]*(xy_overlap[1]))
nx_windows = np.int((xspan-nx_buffer)/nx_pix_per_step)
ny_windows = np.int((yspan-ny_buffer)/ny_pix_per_step)
# Initialize a list to append window positions to
window_list = []
# Loop through finding x and y window positions
# Note: you could vectorize this step, but in practice
# you'll be considering windows one by one with your
# classifier, so looping makes sense
for ys in range(ny_windows):
for xs in range(nx_windows):
# Calculate window position
startx = xs*nx_pix_per_step + x_start_stop[0]
endx = startx + xy_window[0]
starty = ys*ny_pix_per_step + y_start_stop[0]
endy = starty + xy_window[1]
# Append window position to list
window_list.append(((startx, starty), (endx, endy)))
# Return the list of windows
return window_list
# Define a function to draw bounding boxes
def draw_boxes(img, bboxes, color=(0, 0, 255), thick=6):
# Make a copy of the image
imcopy = np.copy(img)
# Iterate through the bounding boxes
for bbox in bboxes:
# Draw a rectangle given bbox coordinates
cv2.rectangle(imcopy, bbox[0], bbox[1], color, thick)
# Return the image copy with boxes drawn
return imcopy
Cell 3: Function to extract features of a single image window¶
In [3]:
# Define a function to extract features from a single image window
# This function is very similar to extract_features()
# just for a single image rather than list of images
def single_img_features(img, color_space='RGB', spatial_size=(32, 32),
hist_bins=32, orient=9,
pix_per_cell=8, cell_per_block=2, hog_channel=0,
spatial_feat=True, hist_feat=True, hog_feat=True):
#1) Define an empty list to receive features
img_features = []
#2) Apply color conversion if other than 'RGB'
if color_space != 'RGB':
if color_space == 'HSV':
feature_image = cv2.cvtColor(img, cv2.COLOR_RGB2HSV)
elif color_space == 'LUV':
feature_image = cv2.cvtColor(img, cv2.COLOR_RGB2LUV)
elif color_space == 'HLS':
feature_image = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)
elif color_space == 'YUV':
feature_image = cv2.cvtColor(img, cv2.COLOR_RGB2YUV)
elif color_space == 'YCrCb':
feature_image = cv2.cvtColor(img, cv2.COLOR_RGB2YCrCb)
else: feature_image = np.copy(img)
#3) Compute spatial features if flag is set
if spatial_feat == True:
spatial_features = bin_spatial(feature_image, size=spatial_size)
#4) Append features to list
img_features.append(spatial_features)
#5) Compute histogram features if flag is set
if hist_feat == True:
hist_features = color_hist(feature_image, nbins=hist_bins)
#6) Append features to list
img_features.append(hist_features)
#7) Compute HOG features if flag is set
if hog_feat == True:
if hog_channel == 'ALL':
hog_features = []
for channel in range(feature_image.shape[2]):
hog_features.extend(get_hog_features(feature_image[:,:,channel],
orient, pix_per_cell, cell_per_block,
vis=False, feature_vec=True))
else:
hog_features = get_hog_features(feature_image[:,:,hog_channel], orient,
pix_per_cell, cell_per_block, vis=False, feature_vec=True)
#8) Append features to list
img_features.append(hog_features)
#9) Return concatenated array of features
return np.concatenate(img_features)
Cell 5: Load images¶
In [4]:
# Read in cars and notcars
cars = []
notcars = []
car_images = glob.glob('vehicles/*/*.png')
for image in car_images:
cars.append(image)
print(len(cars))
nocar_images = glob.glob('non-vehicles/*/*.png')
for image in nocar_images:
notcars.append(image)
print(len(notcars))
# Reduce the sample size because
# The quiz evaluator times out after 13s of CPU time
sample_size = 8790
cars = cars[0:sample_size]
notcars = notcars[0:sample_size]
Data Example Plot¶
In [5]:
# Choose a random indices for plotting
car_idx = np.random.randint(0, len(cars))
notcar_idx = np.random.randint(0, len(notcars))
# Read car and non-car image
car_image = mpimg.imread(cars[car_idx])
notcar_image = mpimg.imread(notcars[notcar_idx])
# Plot an example of a vehicle and non-vehicle image
f, (ax1, ax2) = plt.subplots(1, 2, figsize=(8,8))
ax1.imshow(car_image)
ax1.set_title('Car Image', fontsize=15)
ax2.imshow(notcar_image)
ax2.set_title('Not Car Image', fontsize=15)
plt.savefig('output_images/car_and_notcar.png')
Data Example HOG¶
In [6]:
# Choose a random indices for plotting
car_idx = np.random.randint(0, len(cars))
notcar_idx = np.random.randint(0, len(notcars))
# Read car and non-car image
car_image = mpimg.imread(cars[car_idx])
notcar_image = mpimg.imread(notcars[notcar_idx])
gray_car = cv2.cvtColor(car_image, cv2.COLOR_RGB2GRAY)
gray_notcar = cv2.cvtColor(notcar_image, cv2.COLOR_RGB2GRAY)
orient = 12
pix_per_cell = 8
cell_per_block = 2
# Call our function with vis=True to see an image output
features_car, hog_image_car = get_hog_features(gray_car, orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
features_notcar, hog_image_notcar = get_hog_features(gray_notcar, orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# Plot the examples
fig = plt.figure(figsize=(12,12))
plt.subplot(221)
plt.imshow(car_image, cmap='gray')
plt.title('Example Car Image')
plt.subplot(222)
plt.imshow(notcar_image, cmap='gray')
plt.title('Example Non-Car Image')
plt.subplot(223)
plt.imshow(hog_image_car, cmap='gray')
plt.title('HOG Example Car Image')
plt.subplot(224)
plt.imshow(hog_image_notcar, cmap='gray')
plt.title('HOG Example Non-Car Image')
plt.savefig('examples/hog.png')
Different Color Space for HOG¶
In [7]:
# YCrCb
image = convert_color(car_image)
# Call our function with vis=True to see an image output per channel
_, hog_image_car_R = get_hog_features(image[:,:,0], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_G = get_hog_features(image[:,:,1], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_B = get_hog_features(image[:,:,2], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# YUV
image = cv2.cvtColor(car_image, cv2.COLOR_RGB2YUV)
# Call our function with vis=True to see an image output per channel
_, hog_image_car_Y = get_hog_features(image[:,:,0], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_U = get_hog_features(image[:,:,1], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_V = get_hog_features(image[:,:,2], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# HLS
image = cv2.cvtColor(car_image, cv2.COLOR_RGB2HLS)
# Call our function with vis=True to see an image output per channel
_, hog_image_car_H = get_hog_features(image[:,:,0], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_L = get_hog_features(image[:,:,1], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_S = get_hog_features(image[:,:,2], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# Plot the examples
fig = plt.figure(figsize=(14,14))
plt.subplot(341)
plt.imshow(car_image, cmap='gray')
plt.title('Example Car Image')
plt.subplot(342)
plt.imshow(hog_image_car_R, cmap='gray')
plt.title('HOG Example Channel R')
plt.subplot(343)
plt.imshow(hog_image_car_G, cmap='gray')
plt.title('HOG Example Channel G')
plt.subplot(344)
plt.imshow(hog_image_car_B, cmap='gray')
plt.title('HOG Example Channel B')
plt.subplot(346)
plt.imshow(hog_image_car_Y, cmap='gray')
plt.title('HOG Example Channel Y')
plt.subplot(347)
plt.imshow(hog_image_car_U, cmap='gray')
plt.title('HOG Example Channel U')
plt.subplot(348)
plt.imshow(hog_image_car_V, cmap='gray')
plt.title('HOG Example Channel V')
plt.subplot(3,4,10)
plt.imshow(hog_image_car_H, cmap='gray')
plt.title('HOG Example Channel H')
plt.subplot(3,4,11)
plt.imshow(hog_image_car_L, cmap='gray')
plt.title('HOG Example Channel L')
plt.subplot(3,4,12)
plt.imshow(hog_image_car_S, cmap='gray')
plt.title('HOG Example Channel S')
plt.savefig('examples/hog2.png')
Different Parameters for HOG¶
In [8]:
# YCrCb
image = convert_color(car_image)
# Parameter set 1
orient = 12
pix_per_cell = 8
cell_per_block = 2
# Call our function with vis=True to see an image output per channel
_, hog_image_car_A1 = get_hog_features(image[:,:,0], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_A2 = get_hog_features(image[:,:,1], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_A3 = get_hog_features(image[:,:,2], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# Parameter set 2
orient = 12
pix_per_cell = 8
cell_per_block = 4
# Call our function with vis=True to see an image output per channel
_, hog_image_car_B1 = get_hog_features(image[:,:,0], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_B2 = get_hog_features(image[:,:,1], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_B3 = get_hog_features(image[:,:,2], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# Parameter set 3
orient = 12
pix_per_cell = 4
cell_per_block = 2
# Call our function with vis=True to see an image output per channel
_, hog_image_car_C1 = get_hog_features(image[:,:,0], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_C2 = get_hog_features(image[:,:,1], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
_, hog_image_car_C3 = get_hog_features(image[:,:,2], orient,
pix_per_cell, cell_per_block,
vis=True, feature_vec=False)
# Plot the examples
fig = plt.figure(figsize=(14,14))
plt.subplot(341)
plt.imshow(car_image, cmap='gray')
plt.title('Example Car Image')
plt.subplot(342)
plt.imshow(hog_image_car_A1, cmap='gray')
plt.title('HOG Example Channel 1')
plt.subplot(343)
plt.imshow(hog_image_car_A2, cmap='gray')
plt.title('HOG Example Channel 2')
plt.subplot(344)
plt.imshow(hog_image_car_A3, cmap='gray')
plt.title('HOG Example Channel 3')
plt.subplot(346)
plt.imshow(hog_image_car_B1, cmap='gray')
plt.title('HOG Example Channel 1')
plt.subplot(347)
plt.imshow(hog_image_car_B2, cmap='gray')
plt.title('HOG Example Channel 2')
plt.subplot(348)
plt.imshow(hog_image_car_B3, cmap='gray')
plt.title('HOG Example Channel 3')
plt.subplot(3,4,10)
plt.imshow(hog_image_car_C1, cmap='gray')
plt.title('HOG Example Channel 1')
plt.subplot(3,4,11)
plt.imshow(hog_image_car_C2, cmap='gray')
plt.title('HOG Example Channel 2')
plt.subplot(3,4,12)
plt.imshow(hog_image_car_C3, cmap='gray')
plt.title('HOG Example Channel 3')
plt.savefig('examples/hog3.png')
Cell 4: Search qualified windows in windows_list (need to calculate hog multiple times)¶
In [9]:
# Define a function you will pass an image
# and the list of windows to be searched (output of slide_windows())
def search_windows(img, windows, clf, scaler, color_space='RGB',
spatial_size=(32, 32), hist_bins=32,
hist_range=(0, 256), orient=9,
pix_per_cell=8, cell_per_block=2,
hog_channel=0, spatial_feat=True,
hist_feat=True, hog_feat=True):
#1) Create an empty list to receive positive detection windows
on_windows = []
#2) Iterate over all windows in the list
for window in windows:
#3) Extract the test window from original image
test_img = cv2.resize(img[window[0][1]:window[1][1], window[0][0]:window[1][0]], (64, 64))
#4) Extract features for that window using single_img_features()
features = single_img_features(test_img, color_space=color_space,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
#5) Scale extracted features to be fed to classifier
test_features = scaler.transform(np.array(features).reshape(1, -1))
#6) Predict using your classifier
prediction = clf.predict(test_features)
#7) If positive (prediction == 1) then save the window
if prediction == 1:
on_windows.append(window)
#8) Return windows for positive detections
return on_windows
Cell 4.1 Hog Sub-sampling Window Search¶
In [10]:
# Define a single function that can extract features using hog sub-sampling and make predictions
def find_cars(img,ystart, ystop, scale, svc, X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins):
#1) Create an empty list to receive positive detection windows
on_windows = []
draw_img = np.copy(img)
img = img.astype(np.float32)/255
img_tosearch = img[ystart:ystop,:,:]
ctrans_tosearch = convert_color(img_tosearch, conv='RGB2YCrCb')
if scale != 1:
imshape = ctrans_tosearch.shape
ctrans_tosearch = cv2.resize(ctrans_tosearch, (np.int(imshape[1]/scale), np.int(imshape[0]/scale)))
ch1 = ctrans_tosearch[:,:,0]
ch2 = ctrans_tosearch[:,:,1]
ch3 = ctrans_tosearch[:,:,2]
# Define blocks and steps as above
nxblocks = (ch1.shape[1] // pix_per_cell) - cell_per_block + 1
nyblocks = (ch1.shape[0] // pix_per_cell) - cell_per_block + 1
nfeat_per_block = orient*cell_per_block**2
# 64 was the orginal sampling rate, with 8 cells and 8 pix per cell
window = 64
nblocks_per_window = (window // pix_per_cell) - cell_per_block + 1
cells_per_step = 2 # Instead of overlap, define how many cells to step
nxsteps = (nxblocks - nblocks_per_window) // cells_per_step + 1
nysteps = (nyblocks - nblocks_per_window) // cells_per_step + 1
# Compute individual channel HOG features for the entire image
hog1 = get_hog_features(ch1, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog2 = get_hog_features(ch2, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog3 = get_hog_features(ch3, orient, pix_per_cell, cell_per_block, feature_vec=False)
for xb in range(nxsteps):
for yb in range(nysteps):
ypos = yb*cells_per_step
xpos = xb*cells_per_step
# Extract HOG for this patch
hog_feat1 = hog1[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat2 = hog2[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat3 = hog3[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_features = np.hstack((hog_feat1, hog_feat2, hog_feat3))
xleft = xpos*pix_per_cell
ytop = ypos*pix_per_cell
# Extract the image patch
subimg = cv2.resize(ctrans_tosearch[ytop:ytop+window, xleft:xleft+window], (64,64))
# Get color features
spatial_features = bin_spatial(subimg, size=spatial_size)
hist_features = color_hist(subimg, nbins=hist_bins)
# Scale features and make a prediction
test_features = X_scaler.transform(np.hstack((spatial_features, hist_features, hog_features)).reshape(1, -1))
#test_features = X_scaler.transform(np.hstack((shape_feat, hist_feat)).reshape(1, -1))
test_prediction = svc.predict(test_features)
if test_prediction == 1:
xbox_left = np.int(xleft*scale)
ytop_draw = np.int(ytop*scale)
win_draw = np.int(window*scale)
on_window = ((xbox_left, ytop_draw+ystart), (xbox_left+win_draw,ytop_draw+win_draw+ystart))
on_windows.append(on_window)
cv2.rectangle(draw_img,(xbox_left, ytop_draw+ystart),(xbox_left+win_draw,ytop_draw+win_draw+ystart),(0,0,255),6)
return draw_img,on_windows
Cell 6: Parameters¶
In [11]:
### TODO: Tweak these parameters and see how the results change.
color_space = 'YCrCb' # Can be RGB, HSV, LUV, HLS, YUV, YCrCb
orient = 9 # HOG orientations
pix_per_cell = 8 # HOG pixels per cell
cell_per_block = 2 # HOG cells per block
hog_channel = "ALL" # Can be 0, 1, 2, or "ALL"
spatial_size = (32, 32) # Spatial binning dimensions
hist_bins = 32 # Number of histogram bins
spatial_feat = True # Spatial features on or off
hist_feat = True # Histogram features on or off
hog_feat = True # HOG features on or off
y_start_stop = [360, 650] # Min and max in y to search in slide_window()
Cell 7: Data preprocess¶
In [12]:
car_features = extract_features(cars, color_space=color_space,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
notcar_features = extract_features(notcars, color_space=color_space,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
# Create an array stack of feature vectors
X = np.vstack((car_features, notcar_features)).astype(np.float64)
# Define the labels vector
y = np.hstack((np.ones(len(car_features)), np.zeros(len(notcar_features))))
# Split up data into randomized training and test sets
rand_state = np.random.randint(0, 100)
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.25, random_state=rand_state)
# Fit a per-column scaler
X_scaler = StandardScaler().fit(X_train)
# Apply the scaler to X
X_train = X_scaler.transform(X_train)
X_test = X_scaler.transform(X_test)
print('Using:',orient,'orientations',pix_per_cell,
'pixels per cell and', cell_per_block,'cells per block')
print('Feature vector length:', len(X_train[0]))
Cell 8 : Training¶
In [21]:
# Use a linear SVC
svc = LinearSVC(C=0.1)
# Check the training time for the SVC
t=time.time()
svc.fit(X_train, y_train)
t2 = time.time()
print(round(t2-t, 2), 'Seconds to train SVC...')
# Check the score of the SVC
print('Test Accuracy of SVC = ', round(svc.score(X_test, y_test), 4))
# Check the prediction time for a single sample
pickle.dump( svc, open( "./svc_pickle.p", "wb" ) )
Cell 8.1: Parameter Tuning¶
In [ ]:
from sklearn.model_selection import GridSearchCV
# Use a linear SVC
parameters = {'C':[0.01,0.1,0.5,1,5,10]}
svr = LinearSVC(max_iter=5000)
clf = GridSearchCV(svr, parameters,cv=5)
# Check the training time for the SVC
t=time.time()
clf.fit(X_train, y_train)
t2 = time.time()
print(round(t2-t, 2), 'Seconds to train SVC...')
# Check the score of the SVC
print('Test Accuracy of SVC = ', round(svc.score(X_test, y_test), 4))
# Best Parameter
print('Best C Value = ',clf.best_params_)
pickle.dump( svc, open( "./svc_pickle.p", "wb" ) )
Cell 9: Prediction using windows search¶
In [22]:
image = mpimg.imread('test_images/test6.jpg')
draw_image = np.copy(image)
image_scaled = image.astype(np.float32)/255
t=time.time()
windows = slide_window(image_scaled, x_start_stop=[None, None], y_start_stop=y_start_stop,
xy_window=(96, 96), xy_overlap=(0.8, 0.8))
hot_windows = search_windows(image_scaled, windows, svc, X_scaler, color_space=color_space,
spatial_size=spatial_size, hist_bins=hist_bins,
orient=orient, pix_per_cell=pix_per_cell,
cell_per_block=cell_per_block,
hog_channel=hog_channel, spatial_feat=spatial_feat,
hist_feat=hist_feat, hog_feat=hog_feat)
window_img = draw_boxes(draw_image, hot_windows, color=(0, 0, 255), thick=6)
t2 = time.time()
print(round(t2-t, 2), 'Seconds to Search...')
plt.imshow(window_img)
plt.show()
Cell 9.1 Prediction using Hog Sub-sampling Window Search¶
In [23]:
image = mpimg.imread('test_images/test6.jpg')
draw_image = np.copy(image)
ystart = y_start_stop[0]
ystop = y_start_stop[1]
scale = 1.5
t=time.time()
window_img, window_list = find_cars(image, ystart, ystop, scale, svc, X_scaler, orient, pix_per_cell, \
cell_per_block, spatial_size, hist_bins)
t2 = time.time()
print(round(t2-t, 2), 'Seconds to Search...')
plt.imshow(window_img)
plt.show()
Cell 9.3: Multiscale¶
In [85]:
def find_cars_multiscales(image, ystart, ystop, scale_list, svc, X_scaler, orient, pix_per_cell, \
cell_per_block, spatial_size, hist_bins):
draw_img = np.copy(image)
#1) Create an empty list to receive positive detection windows
on_windows = []
img = image.astype(np.float32)/255
img_tosearch = img[ystart:ystop,:,:]
ctrans_tosearch = convert_color(img_tosearch, conv='RGB2YCrCb')
for scale in scale_list:
if scale != 1:
imshape = ctrans_tosearch.shape
ctrans_tosearch_scaled = cv2.resize(ctrans_tosearch, (np.int(imshape[1]/scale), \
np.int(imshape[0]/scale)))
else:
ctrans_tosearch_scaled = np.copy(ctrans_tosearch)
ch1 = ctrans_tosearch_scaled[:,:,0]
ch2 = ctrans_tosearch_scaled[:,:,1]
ch3 = ctrans_tosearch_scaled[:,:,2]
# Define blocks and steps as above
nxblocks = (ch1.shape[1] // pix_per_cell) - cell_per_block + 1
nyblocks = (ch1.shape[0] // pix_per_cell) - cell_per_block + 1
nfeat_per_block = orient*cell_per_block**2
# 64 was the orginal sampling rate, with 8 cells and 8 pix per cell
window = 64
nblocks_per_window = (window // pix_per_cell) - cell_per_block + 1
cells_per_step = 2 # Instead of overlap, define how many cells to step
nxsteps = (nxblocks - nblocks_per_window) // cells_per_step + 1
nysteps = (nyblocks - nblocks_per_window) // cells_per_step + 1
# Compute individual channel HOG features for the entire image
hog1 = get_hog_features(ch1, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog2 = get_hog_features(ch2, orient, pix_per_cell, cell_per_block, feature_vec=False)
hog3 = get_hog_features(ch3, orient, pix_per_cell, cell_per_block, feature_vec=False)
for xb in range(nxsteps):
for yb in range(nysteps):
ypos = yb*cells_per_step
xpos = xb*cells_per_step
# Extract HOG for this patch
hog_feat1 = hog1[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat2 = hog2[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_feat3 = hog3[ypos:ypos+nblocks_per_window, xpos:xpos+nblocks_per_window].ravel()
hog_features = np.hstack((hog_feat1, hog_feat2, hog_feat3))
xleft = xpos*pix_per_cell
ytop = ypos*pix_per_cell
# Extract the image patch
subimg = cv2.resize(ctrans_tosearch_scaled[ytop:ytop+window, xleft:xleft+window], (64,64))
# Get color features
spatial_features = bin_spatial(subimg, size=spatial_size)
hist_features = color_hist(subimg, nbins=hist_bins)
# Scale features and make a prediction
test_features = X_scaler.transform(np.hstack((spatial_features, hist_features, hog_features)).reshape(1, -1))
#test_features = X_scaler.transform(np.hstack((shape_feat, hist_feat)).reshape(1, -1))
test_prediction = svc.predict(test_features)
if test_prediction == 1:
xbox_left = np.int(xleft*scale)
ytop_draw = np.int(ytop*scale)
win_draw = np.int(window*scale)
on_window = ((xbox_left, ytop_draw+ystart), (xbox_left+win_draw,ytop_draw+win_draw+ystart))
on_windows.append(on_window)
cv2.rectangle(draw_img,(xbox_left, ytop_draw+ystart),(xbox_left+win_draw,ytop_draw+\
win_draw+ystart),(0,0,255),6)
return draw_img,on_windows
test_images_path = glob.glob('test_images/test*.jpg')
test_images = []
for test_image_path in test_images_path:
test_image = mpimg.imread(test_image_path)
test_images.append(test_image)
ystart = y_start_stop[0]
ystop = y_start_stop[1]
scale = [1,1.5,2]
window_img1, window_list1 = find_cars_multiscales(test_images[0], ystart, ystop, scale, svc, \
X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins)
window_img2, window_list2 = find_cars_multiscales(test_images[1], ystart, ystop, scale, svc, \
X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins)
window_img3, window_list3 = find_cars_multiscales(test_images[2], ystart, ystop, scale, svc, \
X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins)
window_img4, window_list4 = find_cars_multiscales(test_images[3], ystart, ystop, scale, svc, \
X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins)
window_img5, window_list5 = find_cars_multiscales(test_images[4], ystart, ystop, scale, svc, \
X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins)
window_img6, window_list6 = find_cars_multiscales(test_images[5], ystart, ystop, scale, svc, \
X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins)
# Plot the examples
fig = plt.figure(figsize=(18,18))
plt.subplot(321)
plt.imshow(window_img1, cmap='gray')
plt.subplot(322)
plt.imshow(window_img2, cmap='gray')
plt.subplot(323)
plt.imshow(window_img3, cmap='gray')
plt.subplot(324)
plt.imshow(window_img4, cmap='gray')
plt.subplot(325)
plt.imshow(window_img5, cmap='gray')
plt.subplot(326)
plt.imshow(window_img6, cmap='gray')
plt.savefig('examples/detecttest_1.png')
Cell 10: Multiple Detections & False Positives¶
In [86]:
from scipy.ndimage.measurements import label
def add_heat(heatmap, bbox_list):
# Iterate through list of bboxes
for box in bbox_list:
# Add += 1 for all pixels inside each bbox
# Assuming each "box" takes the form ((x1, y1), (x2, y2))
heatmap[box[0][1]:box[1][1], box[0][0]:box[1][0]] += 1
# Return updated heatmap
return heatmap# Iterate through list of bboxes
def apply_threshold(heatmap, threshold):
# Zero out pixels below the threshold
heatmap[heatmap <= threshold] = 0
# Return thresholded map
return heatmap
def draw_labeled_bboxes(img, labels):
# Iterate through all detected cars
for car_number in range(1, labels[1]+1):
# Find pixels with each car_number label value
nonzero = (labels[0] == car_number).nonzero()
# Identify x and y values of those pixels
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Define a bounding box based on min/max x and y
bbox = ((np.min(nonzerox), np.min(nonzeroy)), (np.max(nonzerox), np.max(nonzeroy)))
# Draw the box on the image
cv2.rectangle(img, bbox[0], bbox[1], (0,0,255), 6)
# Return the image
return img
In [92]:
heat = np.zeros_like(image[:,:,0]).astype(np.float)
# Add heat to each box in box list
heat = add_heat(heat,window_list)
# Apply threshold to help remove false positives
heat = apply_threshold(heat,2)
# Visualize the heatmap when displaying
heatmap = np.clip(heat, 0, 255)
# Find final boxes from heatmap using label function
labels = label(heatmap)
draw_img = draw_labeled_bboxes(np.copy(image), labels)
fig = plt.figure(figsize=(14,14))
plt.subplot(121)
plt.imshow(draw_img)
plt.title('Car Positions')
plt.subplot(122)
plt.imshow(heatmap, cmap='hot')
plt.title('Heat Map')
fig.tight_layout()
plt.savefig('examples/heatmap.png')
Cell 11: Video Pipeline¶
In [93]:
def pipeline(image,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins,\
ystart,ystop,scale_list = [1,2]):
draw_image = np.copy(image)
window_img, window_list = find_cars_multiscales(image, ystart, ystop, scale_list, svc, X_scaler, \
orient, pix_per_cell, cell_per_block, spatial_size, hist_bins)
heat = np.zeros_like(image[:,:,0]).astype(np.float)
# Add heat to each box in box list
heat = add_heat(heat,window_list)
# Apply threshold to help remove false positives
heat = apply_threshold(heat,2)
# Visualize the heatmap when displaying
heatmap = np.clip(heat, 0, 255)
# Find final boxes from heatmap using label function
labels = label(heatmap)
draw_img = draw_labeled_bboxes(np.copy(image), labels)
return draw_img
In [94]:
test_images_path = glob.glob('test_images/test*.jpg')
test_images = []
for test_image_path in test_images_path:
test_image = mpimg.imread(test_image_path)
test_images.append(test_image)
image1 = test_images[0]
image_draw1 = pipeline(image1,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, \
hist_bins,ystart,ystop,scale_list = scale)
image2 = test_images[1]
image_draw2 = pipeline(image2,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, \
hist_bins,ystart,ystop,scale_list = scale)
image3 = test_images[2]
image_draw3 = pipeline(image3,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, \
hist_bins,ystart,ystop,scale_list = scale)
image4 = test_images[3]
image_draw4 = pipeline(image4,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, \
hist_bins,ystart,ystop,scale_list = scale)
image5 = test_images[4]
image_draw5 = pipeline(image5,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, \
hist_bins,ystart,ystop,scale_list = scale)
image6 = test_images[5]
image_draw6 = pipeline(image6,svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, \
hist_bins,ystart,ystop,scale_list = scale)
In [95]:
# Plot the examples
fig = plt.figure(figsize=(18,18))
plt.subplot(321)
plt.imshow(image_draw1, cmap='gray')
plt.subplot(322)
plt.imshow(image_draw2, cmap='gray')
plt.subplot(323)
plt.imshow(image_draw3, cmap='gray')
plt.subplot(324)
plt.imshow(image_draw4, cmap='gray')
plt.subplot(325)
plt.imshow(image_draw5, cmap='gray')
plt.subplot(326)
plt.imshow(image_draw6, cmap='gray')
plt.savefig('examples/detecttest.png')
In [96]:
from moviepy.editor import VideoFileClip
def VideoPipeline(inputVideo, outputVideo,svc,X_scaler, orient, pix_per_cell, cell_per_block, \
spatial_size, hist_bins,ystart,ystop,scale_list):
myclip = VideoFileClip(inputVideo)
clip = myclip.fl_image(lambda image: pipeline(image,svc,X_scaler, orient, pix_per_cell, \
cell_per_block, spatial_size, hist_bins,ystart,ystop,scale_list))
clip.write_videofile(outputVideo, audio=False)
VideoPipeline('test_video.mp4', 'test_video_result.mp4',svc,X_scaler, orient, pix_per_cell, \
cell_per_block, spatial_size, hist_bins,ystart,ystop,scale_list = scale)
In [97]:
VideoPipeline('project_video.mp4', 'project_video_result.mp4',svc,X_scaler, orient, pix_per_cell, cell_per_block, spatial_size, hist_bins,ystart,ystop,scale_list = scale)
In [ ]: